تطبيق نموذج YOLOv8 في الكشف عن السفن من الصور الفضائية لدعم أنظمة المراقبة البحرية في الزمن الحقيقي

Abstract

• This study evaluates the effectiveness of the state-of-the-art YOLOv8 model for ship detection in high-resolution satellite imagery. The model was trained on a dataset of 2,500 satellite images encompassing ships in diverse environmental conditions. The data was split into 70% for training, 15% for validation, and 15% for testing. Results demonstrate the superiority of the proposed model, achieving a mean Average Precision (mAP@0.5) of 90.5%, Precision of 91.3%, and Recall of 89.7%. A MIL Tracker algorithm was integrated to enhance monitoring continuity across image sequences, reducing ID switches by 35%. The model showed significant superiority in speed and accuracy compared to YOLOv5 and Faster R-CNN models under the same testing conditions, making it a practical solution for real-time maritime monitoring applications.